A SEPAR Grant Will Help Advance an AI Model to Predict Asthma Exacerbations

Predicting which patients with asthma are at greater risk of experiencing an exacerbation and gaining a better understanding of the mechanisms involved in each episode remain major challenges in managing the disease. To address this need, Dr. Astrid Crespo, a researcher with the Chronic Respiratory Diseases Research Group at the Sant Pau Research Institute (IR Sant Pau) and a physician in the Department of Pulmonology and Allergy at Hospital de Sant Pau, is leading a project to develop a predictive model integrating artificial intelligence, metabolic biomarkers, microbiology, and medical imaging.

The project has received one of the 2026 Research Time Grants awarded by the Spanish Society of Pulmonology and Thoracic Surgery (SEPAR) in the senior category. The grant will enable Dr. Crespo to devote more time to research and support progress toward a more personalized approach to preventing and treating asthma attacks.

Titled “Development of an AI-Based Predictive Model to Identify Metabolic Biomarkers in Asthma Exacerbations: A Multidimensional Approach,” the project will combine different sources of information to identify patterns associated with the risk, frequency, and severity of these episodes.

“This grant allows us to devote more time to a project that stems from a very specific clinical need: identifying in advance which patients are at greater risk of experiencing an exacerbation and better understanding the mechanisms involved in each episode,” explains Dr. Astrid Crespo.

A Clinical Problem That Is Difficult to Predict

Exacerbations are episodes of acute worsening of asthma that may require changes in treatment, emergency care, or hospitalization. In addition to affecting the quality of life of people with asthma, they place a substantial burden on healthcare services and can make the disease more difficult to control.

However, not all exacerbations have the same cause or are driven by the same inflammatory mechanisms. This heterogeneity makes it difficult to determine which patients will experience another attack, when it may occur, and how severe it will be.

The biomarkers currently used, when analyzed individually, have limited predictive capacity. The project will therefore examine different dimensions of the disease together to obtain a more comprehensive characterization of each patient during an exacerbation.

The study will also examine the differences between exacerbations with an eosinophilic inflammatory profile and those with a non-eosinophilic profile. The former involve elevated levels of eosinophils, a type of white blood cell involved in certain inflammatory and allergic responses. The latter involve other mechanisms that may be related, for example, to infections, irritants, or other inflammatory pathways.

Distinguishing between these two profiles is important because they may respond differently to treatment and require different therapeutic strategies. The project will investigate whether certain metabolic biomarkers can differentiate between them during an exacerbation and provide information about disease progression.

A Multidimensional View of Asthma Attacks

The project will combine metabolic profiles found in the blood, sputum microbiology, and information obtained through chest computed tomography. The aim is to examine the relationships between these data sources and identify signals that can provide a more precise characterization of each exacerbation.

Metabolic analysis will make it possible to identify small molecules in the blood whose behavior may vary depending on the type of inflammation, disease severity, or risk of experiencing further episodes. At the same time, microbiological analysis of sputum will help determine whether microorganisms are present in the airways and how they may be related to asthma attacks.

These data will be supplemented with information extracted from medical images using radiomics, a methodology that converts radiological scans into a large set of quantitative data on tissue characteristics and lung abnormalities. These variables may reveal patterns that are not always apparent through conventional visual interpretation.

The team will also examine the relationship between these biomarkers and associated conditions such as bronchiectasis, bronchial hypersecretion, and chronic bronchial infection. In this way, the research seeks to deepen understanding of the mechanisms involved in different types of exacerbations and explain why their progression may vary from one patient to another.

“The main innovation is that we will not analyze each piece of data in isolation. We want to integrate metabolic, microbiological, and radiological information to better understand the mechanisms underlying each exacerbation and obtain a more comprehensive view of every patient,” the researcher notes.

Artificial Intelligence to Integrate Complex Data

The large volume and diversity of information generated will be integrated using machine learning algorithms capable of simultaneously analyzing multiple variables and exploring complex relationships between them. Based on this analysis, the researchers will work to develop and validate a predictive model designed to identify patterns associated with exacerbations. Artificial intelligence may facilitate the detection of combinations of variables that would otherwise go unnoticed using conventional approaches or when each biomarker is examined separately.

“Artificial intelligence gives us the opportunity to connect very different types of data and identify patterns that would be difficult to detect using conventional methods. The aim is not to replace clinical assessment, but to complement it with information that can help us make more precise decisions,” says Dr. Crespo.

The study is designed as a prospective, longitudinal investigation involving people with asthma and will include the collection of clinical data, biological samples, and radiological images. The research time grant will facilitate the coordination of these activities and support patient recruitment, sample processing, and the integration of results.

Toward More Personalized Follow-Up Care

The early identification of patients at greater risk could help tailor the intensity of follow-up care, strengthen preventive measures, and review treatment before another attack occurs. The model could also support clinical decision-making based on each person’s inflammatory, metabolic, microbiological, and radiological characteristics.

The project therefore contributes to progress toward more predictive and personalized respiratory medicine, in which treatment and follow-up care are based not only on general parameters but also on a more detailed understanding of the mechanisms involved in each patient’s disease.

“The ultimate goal is not only to predict an exacerbation, but also to generate knowledge and tools that are useful in clinical practice. Being able to anticipate these episodes would help us intervene earlier, better personalize follow-up care and treatment, and reduce the impact these attacks have on the lives of people with asthma,” concludes Dr. Crespo.